[ https://issues.jboss.org/browse/ISPN-3633?page=com.atlassian.jira.plugin.... ] William Burns commented on ISPN-3633: ------------------------------------- I was finally able to reproduce the problem. It has to be done as so. A - primary owner C - non owner # C gets value for k from A the current value and caches it in L1 # A is updating k with a new value in an explicit tx along with another entry that maps to C # A sends and completes prepare to C where it believes it is only updating the other entry (note commit is not yet sent) # State transfer starts and now C is the owner of k (note data has not been transferred yet) # A now commits the tx which causes an invalidation to be sent to C and the value to be marked as modified # State transfer finally is sent to C for data and is ignored. Actually thinking about this it may have a worse issue even when L1 is not enabled that the data cached is the old value as it wouldn't retrieve the value from the put command. I would need to test that still though. -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira